The Real Story Behind Prozac's Approval by the FDA

(This is taken from Chapter 4 of Peter Breggin's book, Talking Back to Prozac.)

Most people seem to believe that the FDA conducts its own independent studies of drugs and then decides whether or not to approve them. Nothing could be further from the truth. The FDA, in fact, doesn't have any money to perform its own studies during the approval process for drugs.

All FDA drug studies are constructed, supervised, and paid for by the drug companies themselves, using doctors and research teams of their own choosing - often people with long-established relationships with the company. It seems obvious, but should be underscored, that the pharmaceutical companies do everything they can to make their studies turn out right.

Regulations require that a new drug must prove its efficacy in double-blind controlled studies comparing it to placebo and to other drugs of established efficacy. Double-blind means that neither the doctors nor the patients know who is getting what kind of pill. Placebo-controlled means that some patients in a comparison control group will be given an inactive substance in pill form - the placebo, or sugar pill. The placebo is, in effect, a fake drug.

If, for example, Prozac does not perform any better than the placebo, Prozac will be seen as ineffective.

Placebo plays a key role in scientific drug studies because it's been repeatedly demonstrated that up to 50 percent or more of depressed patients improve on the sugar pill. In some studies, nearly 90 percent have improved on placebo.

The FDA allowed Lilly to use the "placebo washout," a very questionable but commonly used technique in drug studies . All patients were started off on placebo for approximately one week (4-14 days). Those patients who showed improvement on placebo were then dropped from the study and the trials were begun all over again from scratch with a placebo and a drug group.

Using the placebo washout helps make a drug seem more effective than it is. For example, some of those patients "washed out" of the testing because they responded positively to the placebo might not have reacted positively to the drug if they had received it in the second part of the testing. That is, if they had not been dropped from the actual trials, the placebo responders might have once again reacted positively to the sugar pill but not to the drug with its frequently unpleasant side effects. Even if the placebo washout reduced the number of positive responders equally in both the sugar pill and the drug groups, this reduction in the total number of positive responders in both groups would favor the drug. Why? Statistically, the same difference between two smaller groups is more significant than. the same difference between two larger groups. The placebo washout purposely produces an unnatural pool of patients. It is unscientific.

Before the FDA approves a psychiatric drug, it typically requires that the drug company produce two or more research protocols that demonstrate significant efficacy for the drug. Each protocol has a specific set of rules developed by the drug company. In the case of Prozac, one protocol included ten separate studies under different leadership at different sites.

The results were then pooled to make one pool of data for statistical purposes. The total number of individual studies or research projects in the various Prozac protocols varied from one to ten.

A typical Lilly protocol for Prozac in its FDA approval process required randomly dividing a group of depressed patients into two similar sections. One section was given placebo for four to six weeks and the other was given Prozac for the same period of time. In other protocols, the group was divided into three parts: one taking placebo; one taking Prozac; and one taking an older, proven antidepressant for comparison.

The protocols used a variety of tests to evaluate week-by-week improvement. Some of these involved self-assessment, but most entailed brief interviews with professionals who checked off symptoms lists. No intensive interviews were utilized.

Each individual study was directed by a principle investigator, a psychiatrist selected by Eli Lilly to ensure that the project was conducted according to the principles laid down by the drug company. Some studies were carried out at universities and others at private research firms that specialize in performing drug company sponsored research.

How the Subjects Were Selected

Potential subjects were interviewed by investigators in each study to determine whether or not they met the standards for major depression as defined in the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-IIIR).⁽¹⁾ To qualify for major depression, the individual must show signs of depressed mood or loss of pleasure or interest in life. These are the first two items on a list of nine items, and the individual must suffer from at least a total of five. The other criteria include the following: significant weight change; sleep disturbance; psychological and physical agitation or retardation; fatigue or loss of energy nearly every day; feelings of worthlessness or "excessive or inappropriate guilt"; indecisiveness and other signs of difficulty thinking and concentrating; and recurrent thoughts of death or suicide.

Most people believe that when Prozac was approved for depression, it had been thoroughly tried on extremely depressed patients and had proven life-saving.

In actuality, the Prozac studies as designed by Lilly excluded all patients with serious tendencies toward suicide. This deliberate exclusion was part of the formal protocol, or organization, of each study used for FDA approval in the United States.⁽²⁾ Advocates of the drug would wholly overlook this in their enthusiastic reviews, giving the false impression that Prozac is a potentially life-saving drug. No antidepressant has ever been shown to prevent suicide, and Eli Lilly apparently didn't want to risk finding out whether Prozac would also fail to prevent suicide.

Hospitalized psychiatric patients were also excluded from nearly all of the studies, including every one that was used to approve the drug.

There were no children or elderly adults in the Lilly sponsored FDA studies of Prozac. Once any drug is approved for marketing by the FDA, however, there is nothing to stop psychiatrists from prescribing it for these vulnerable groups. Prozac, in fact, is being widely recommended for children and youth. One such recommendation is in a recent popular book by Columbia University professor of psychiatry Ronald Fieve. Fieve states, "although scientific research in this area is scanty and incomplete the evidence so far indicates that children and adolescents can safely be given Prozac. . . . " The potentially tragic consequences of this practice are underscored in the findings of the government's General Accounting Office (GAO) investigation of the FDA - that children are especially likely to fall victim to adverse reactions that slip through the FDA premarketing tests of drugs.

How Large And How Long Were The Studies

Misleading totals for patients tested in clinical trials are common. In Everything You Need to Know About Prozac (1991), psychiatrist Jeffrey Jonas from Fair Oaks Hospital in Summit, New Jersey,⁽³⁾ and medical writer Ron Schaumburg defended Prozac by stating, "Some of the reassurance comes from the data on over 11,000 patients who took the drug in clinical trials."

The 11,000 figure appears in an August 31, 1990, "Dear Doctor" letter written by Eli Lilly to American physicians to counteract concerns that Prozac could increase suicidality. The letter states, "More than 11,000 individuals participated in clinical trials for Prozac, including over 6,000 treated with Prozac." Jonas and Schaumburg must have misread this statement, since it indicates that the drug was actually taken by "over 6,000" patients rather than 11,000 patients. However, when a drug company bandies about large - if relatively meaningless - numbers, the numbers are likely to be misread or misinterpreted in favor of the company. The figure of 6,000 patients, meanwhile, is itself potentially misleading.

The drug label as reprinted in the Physicians' Desk Reference (PDR) states that there were "5,600 Prozac exposed" individuals in the premarketing period of evaluation.⁽⁴⁾ But this apparently includes patients given the drug under a variety of conditions other than actual clinical trials. In another place in the PDR, it is stated that 4,000 patients received Prozac in "US premarketing clinical trials." But most of these patients, apparently, were not in placebo-controlled studies - the only ones relevant for efficacy studies. A table of adverse reactions in the PDR reviews 1,730 patients who were involved in "placebo-controlled clinical trials."

When we focus on the actual numbers of patients in the trials used by the FDA in the drug approval process, the totals shrink much further. Only three favorable protocols, involving seventeen studies and several hundred patients, were found scientifically adequate enough by the FDA to use them as evidence for approving the drug. Using material obtained through the Freedom of Information Act, I went through each of these seventeen studies, one by one, to add up the number of Prozac patients who actually completed the four-, five-, or six-week trials used as the basis for FDA approval. The grand total turned out to be a meager 286 patients.

It is astonishing to realize that the approval of Prozac was based on fewer than 300 patients culled, by various means that we will examine, from the original cast of thousands. It is safe to guess that few if any physicians or patients who rely on FDA studies have any idea that the actual number of Prozac patients completing the trials was so small.

Most people imagine that patients in the FDA approval studies take the test drugs for months and years on end before the drug can be approved, but as we've already pointed out, the scientifically controlled trials for Prozac lasted only a few weeks. An occasional less scientifically rigorous project lasted longer, but 86 percent of all the patients in all the studies were treated for "three months or less." Only 63 patients were on fluoxetine for a period of more than two years before completion of the premarketing studies and the FDA approval of Prozac. In effect, anyone now taking Prozac for more than a few weeks is part of a giant ongoing experiment on its longer-term effects.

Failures Are Forgiven

According to the FDA approval process, it doesn't matter how many times a drug fails to prove useful in its clinical trials. Innumerable scientific studies can show the drug to be ineffective, but as long as two or more show statistical superiority over placebo, the drug can win approval.

In its "Summary of Basis of Approval," dated October 3, 1988, the FDA states that fourteen protocols involving controlled studies were submitted by Lilly. Four compared Prozac to placebo, and of these, three were used by the FDA as evidence of some beneficial effect. One showed none at all. Of the remaining ten studies, eight showed Prozac to have no positive effect. Overall, there were more negative efforts than positive, but this made no difference in the approval process.

In six out of the seven studies where it was included, imipramine (Tofranil), a very old drug, did better than Prozac. That, too, made no difference in the approval process.

These results do not sound very inspiring, but an examination of the three positive protocols will prove much more disheartening. The analysis that follows was painstakingly garnered from several volumes of FDA data that were decidedly not user-friendly. It is worth giving the reader this rare, and perhaps unprecedented, window into the FDA - A seldom illuminated region of the government.

Protocol 27. Scrambling to Make It

One of the three sets of protocols used to prove the efficacy of Prozac was called Protocol 27. The following information about the protocol is taken from the Food and Drug Administration's March 28, 1985, "Review and Evaluation of Efficacy Data." All page numbers refer to that document.

At the start, Protocol 27 involved more than 700 patients at six separate sites in studies run by different principal investigators. But by the time the entire protocol was completed, and all the data from one of the six sites was excluded, fewer than 150 Prozac patients remained in the whole protocol. Of these, only 104 completed the six-week trials.

The protocol compared three agents: Prozac; an older antidepressant called imipramine (Tofranil); and placebo. The six separate studies, as reported by the drug company and then further analyzed and summarized by the FDA, produced the following results:

1. J. P. Feighner, M.D., from the Feighner Institute in San Diego, according to the FDA found that the older antidepressant, Tofranil, showed significant improvement in the patients in "all variables." However, "fluoxetine was not shown to be consistently different than placebo" (p. 21). In other words, Prozac was a bust.

An April 3,1984, in-house FDA memo by Walter Sloboda, a psychologist in the Division of Scientific Investigation, discussed criticisms of Feighner's studies based on an on-site FDA investigation. Among other things, the FDA found that in Protocol 27 a patient was mistakenly given Prozac in addition to Tofranil and that the error had not been properly recorded. Also, in several cases, a variety of abnormal laboratory findings were ignored, entailing risk to patients. Dr. Feighner, the report said, agreed with these observations by the FDA and promised to remedy them.

In addition, Sloboda's memo discussed a consumer complaint in which a patient alleged that she had been given Prozac in a trial and that this initiated an emotional deterioration resulting in hospitalization and electroshock treatment. We have no information on the outcome of the investigation concerning the patient's complaint.

Feighner's practices in conducting Prozac research were criticized again in an August 7, 1984, letter from the FDA's Frances 0. Kelsey, Ph.D., M.D., Director of the Division of Scientific Investigation, Office of Compliance, Center for Drugs and Biologics. Kelsey found that Feighner "had several departures from Food and Drug Administration regulations or commonly accepted drug investigational practices." Without mentioning specific deviations, the letter emphasized the need for Feighner to follow the inclusion and exclusion criteria of the studies.⁽⁵⁾

2. Jay B. Cohn, M.D., a psychiatrist from the University of California in Los Angeles, produced according to the FDA report "seemingly overwhelmingly positive results." However, the statistical manipulations required to achieve them were scientifically unacceptable. For example, Cohn ended up comparing how the Prozac patients did at six weeks with how the placebo patients did at two weeks. Reluctant to come down too hard on a drug company, the FDA observed, "Hence, this study can, at best, be said to be supportive" (p. 27). In contrast to patients on Prozac, patients on the older antidepressant did show improvement. Another Prozac bust.

Cohn, meanwhile, had been sent an extraordinarily critical letter from Frances 0. Kelsey, Ph.D., M.D., Director, Division of Scientific Investigations of the FDA. The inspection under the FDA's Bioresearch Monitoring Program found that Cohn had failed to indicate that two of the subjects suffered from a "past history of alcoholism" and that a third subject, who had been treated three times for alcoholism in the prior year, had cirrhosis of the liver. It was also found that in an earlier study in the Prozac FDA approval process, Cohn had failed to obtain written informed consent and had "backdated the consent form."

Eventually the FDA discarded the Cohn study as invalid. Nonetheless, in 1985 Cohn and Charles Wilcox, M.B.A., published the study in the Journal of Clinical Psychiatry as part of an Eli Lilly sponsored symposium. They describe their study as an unambiguous success story for Prozac. They give no recognition of the fact that the FDA invalidated and rejected this study. As far as the general medical community knows to this very day, the Cohn study unequivocally proved the efficacy of Prozac.

3. David L. Dunner, M.D., a psychiatrist from the University of Washington in Seattle, found the older antidepressant was effective. But according to the FDA, "There was essentially no difference in efficacy between fluoxetine and placebo" (p. 34). A third failure for Prozac.

4. Bernard 1. Grosser, M.D., from the department of psychiatry of the University of Utah, came up with the same negative result as Dunner. According to the FDA, "imipramine produced significantly more improvement than placebo on all major efficacy variables at endpoint." However, fluoxetine was not shown to be consistently different than, placebo (p. 42). A fourth negative outcome for Prozac.

Grosser's seeming bias in favor of Prozac is disclosed in an October 26, 1984, letter of complaint sent to him by FDA official Frances Kelsey. Among other criticisms, Kelsey observed that Grosser's informed consent form did not conform with regulations because it describes Prozac ". . . as effective or more effective than imipramine. .. ." Kelsey explained, "Since the purpose of the investigational drug studies is to prove the drug's safety and effectiveness, such statements cannot be made while the drug is being evaluated." Interestingly, Grosser's study failed to confirm his bias.

5. F. S. Abuzzahab, Sr., M.D., from the University of Minnesota Department of Psychiatry, again according to the FDA's analysis, showed that Prozac "produced more improvement than placebo on a few variables. The differences, however, were not consistent and included only some key variables" (p. 47). The older antidepressant was no better. This was essentially a fifth bust.

A December 13, 1984, letter from the FDA's Frances Kelsey to Abuzzahab was critical of his practices in Lilly-sponsored Prozac studies. The letter stated, "Objectionable conditions were found to exist in both of the clinical studies for most of the subjects audited." In particular, the inspection found "that you did not adhere to the protocols and that you did not give proper notifications when making protocol changes.

6. James D. Bremner, M.D., a psychiatrist in Olympia, Washington, did find that Prozac, like the older antidepressant, was better than placebo on "most variables," according to the FDA. This was the only positive study among the six for Prozac. The older antidepressant, however, seemed more effective than Prozac, showing improvement on all variables (p. 30). Only 22 patients finished on Prozac. As another important confounding factor, about one-third of the patients received other psychiatric medications - minor tranquilizers and sedatives - in addition to Prozac. There is no way to tell if the results would have been positive if the patients had taken Prozac by itself.

In addition to these individual problems with the studies in Protocol 27 that we have mentioned, all of the protocols came under severe criticism. The FDA's November 13, 1984, "In-House Meeting on Fluoxetine" brought together the leading agency officials concerned with monitoring the Prozac application process, including the division head, Paul Leber. The minutes of the meeting, written by Tony De Cicco, stated, "This Agency has discovered a flaw in the experimental design and execution of the Fluoxetine studies." The flaw was located in "the main efficacy trials," seemingly indicating all the studies we will be examining in this chapter.

According to De Cicco's minutes, patients who were not doing well after the second week of the efficacy studies had the double-blind broken and, if found to be taking placebo or Tofranil, were then switched to Prozac. The blind was then broken at six weeks in patients doing well, in order to continue them on their assigned medication after the study.

The manipulations caused two extreme compromises of all the studies. First, as indicated in the analysis of the Cohn study, the efficacy of Prozac at six weeks ended up being compared to the efficacy of placebo at two weeks - a very "biased" analysis. Second, with the blind broken, investigator bias could compromise the results.

In addition, the "In-House Meeting on Fluoxetine" found that very large dropout rates were impairing the analysis of data. The minutes were especially critical of the seemingly positive results of the Cohn study, but the criticism also applied to all the individual studies in all the main efficacy protocols.

The problems found in these efficacy studies should have invalidated them. But Lilly was allowed to present its data for approval by the FDA.

Despite the bias in favor of Prozac built into the studies by Lilly, at the conclusion of Protocol 27 the picture looked very grim for the drug. But Lilly did not give up. It reshuffled the data, first by removing some but not all of Cohn's embarrassing data. Then Lilly excluded all patients who had received other psychiatric medications along with Prozac. Finally, it pooled the pruned data from the remaining five studies into one batch for statistical analysis, treating them as if they belonged to one study. This increased the total number of subjects for analysis, making it easier to demonstrate statistical significance for relatively minor improvements in the Prozac patients.

According to the FDA's March 28, 1985, efficacy review, the results of this strained effort to pool the data were not convincing: "Imipramine was clearly more effective than placebo, whereas fluoxetine was less consistently better than placebo." The FDA concluded, "This study is supportive but not strongly positive in demonstrating fluoxetine's role in the treatment of depression" (p. 49). Basically, Prozac had been shown, once again, to be a bust.

Undaunted, Lilly reworked the numbers one more time and resubmitted the new calculations to the FDA, now excluding all of the Cohn data as invalid, and again pooling the other five sites. Discarding the Cohn study eliminated 25 percent of the total protocol patients who completed the six-week trials and should have rendered the pooling process invalid; but the FDA accepted it. At the FDA's request, Lilly also reincluded all the patients who had been given additional psychiatric drugs during their Prozac trials. And at last, pay dirt. Lilly managed to come up with a positive result for the pooled studies on four of several measurements of improvement. Based entirely on the reworked figures, and ignoring the multiple flaws and the failure of all but one of the individual studies to demonstrate efficacy, the FDA concluded: "The revised pooling of Protocol 27 can be said to contribute to the judgment of substantial evidence of efficacy" (p. 50A).

Pooling data from separate negative studies in order to get a positive overall result is open to criticism. In fact, the FDA's own regulations on advertising specifically reject the use of such manipulations. The regulations state that drug company ads should not "use statistics based on pooled data from inconclusive studies. These particular pooled studies were worse than "inconclusive"; all but one were outright negative.

Pooling negative studies is questionable enough. Dropping one of the studies in the process, eliminating 25 percent of the Prozac patients who completed the trials, is wholly unacceptable.

Notice the extremes to which Lilly and the FDA had to go to make the numbers work. Notice how much this has to do with statistical juggling and how little with real people. There's no indication that any person actually recovered from major depression, because that wasn't even addressed. Instead, they were tested for relative degrees of improvement compared to the placebo patients as measured on symptom checklists. Although patients had to meet the DSM-III-R criteria for major depression in order to be admitted into the study, the patients were not reinterviewed at the conclusion to see if they had recovered and no longer warranted the diagnosis.

There is evidence that the patients themselves did not feel dramatically improved on Prozac. Of the two rating scales that allowed the patients to record their own impression of the drug, one showed no difference between Prozac and placebo, and one showed some efficacy for the drug.

Could the clinicians have been influenced in their positive judgments of Prozac even though they were not supposed to know which patients were getting the drug? Yes. Prozac patients, as we shall see, have characteristic side effects - such as insomnia, nightmares, anxiety and nervousness, upset stomach, and weight loss - that could have allowed the rating physicians to guess who was on what drug. In particular, the other drug in the study, Tofranil (imipramine), is very sedating, while Prozac is activating. Consciously or unconsciously, accurately guessing who was on the drug could have influenced the ratings bv the doctors.

Earlier we pointed out that studies that fail are forgiven. There's another way this forgiveness takes place within each and every study. Overall in Protocol 27, only slightly more than one-half of the Prozac patients managed to stay on the drug for the whole six weeks. The others dropped out, usually because of adverse drug effects and lack of efficacy. How is it possible to claim, as many psychiatrists have, that 70 to 80 percent of patients benefit from Prozac, when as much as 50 percent of the patients do not even continue taking the drug for a brief six-week trial?

The high dropout rate in these six-week-long studies in part answers a frequently asked question, "Why are the drug studies so short?" In the case of Prozac, if the studies had been even a week or two longer, the trend indicates that the vast majority of the Prozac patients would have dropped out, underscoring the failure of the drug as a therapeutic agent.

In summary, in a multisite protocol in which only one of six studies turned out positive for Prozac, pooled data were shuffled and reshuffled, with much of it eliminated, in order to reach statistical significance on a few measures. These machinations were essentially flawed from the start by conditions such as the placebo washout, the extremely short length of the trials, the very small numbers of Prozac patients who completed the trials, the use of superficial symptom checklists to determine efficacy, the exclusion of suicidal or hospitalized patients, the inclusion of patients on multiple psychiatric drugs, and the better performance of the older antidepressant.

Interestingly, Lilly employees Paul Stark, Ph.D., and C. David Hardison, Ph.D., published the results of the pooled data from the five studies without mentioning the fact that the individual studies failed to show efficacy for Prozac, without mentioning the various FDA criticisms, and without giving the impression that Prozac's performance was at best weak. Unlike the FDA analysis and conclusion, which shows Tofranil to be superior to Prozac, the Lilly version claims Prozac to he as effective as ("comparable to") Tofranil.

Protocol 19: More Doubts

Another of the three key positive Prozac protocols, Protocol 19, was conducted by Louis Fabre, Jr., a Houston, Texas, psychiatrist who frequently does FDA research on behalf of drug companies. According to the FDA's March 28, 1985, review, the placebo washout was again employed, and "serious suicidal risk" was an excluding factor (p. 52). Fabre compared Prozac and placebo. Only 47 patients were entered into the study, 10 were then dropped because they couldn't be properly evaluated, and ultimately only 25 finished the trials. Of these completers, only 11 had been given Prozac (p. 58).

With this slim number of patients completing the protocol, obtaining a positive result involved considerable statistical maneuvering. For example, an additional five Prozac and seven placebo patients were counted in the statistical analysis for efficacy, even though they never finished the trial (p. 58).

The trial, as planned, was only five weeks in duration. But according to De Cicco's critical analysis from the in-house FDA meeting, "Fabre has a 4-week trial at the most."

Considering all this, the study should have been discarded as worthless, yet it became one of the cornerstones for approving Prozac.

How did the Prozac patients in the Fabre protocol rate their own response to the drug? They rated themselves no better than the placebo patients rated themselves. In other words, Prozac was no better than placebo from the patients' viewpoint. These negative results occurred despite the placebo washout, which tends to put placebo at a disadvantage compared to the drug. The same critique of the double-blind - that the doctors probably could tell the drug patients from the placebo patients - applies to this and all the FDA Prozac studies.

Meanwhile, is there any reason to question Fabre's integrity? While no legal actions have been brought against him in regard to his studies on Prozac for Eli Lilly, Fabre has recently been accused in a civil suit of gross misconduct in regard to testing Halcion for the Upjohn Company. He is charged, along with Upjohn, with participation in a "conspiracy, first to market Halcion and keep Halcion ... on the market." In regard to studies he conducted for Upjohn from 1973 to 1975, he is accused of the following:

All of the studies done by Dr. Fabre at the Portland Clinic were falsified. The studies were not "double blind," as they should have been; the drugs were decoded, and both the patient and the investigator knew what drugs the patients were taking.

Dr. Fabre and Upjohn have denied all allegations.

Protocol 62: "Seriously Flawed"

According to the FDA's October 3, 1988, "Summary Basis of Approval," Protocol 62 was considered to be the weakest of the three positive protocols. It was also the largest, initially involving 900 patients at ten centers,⁽⁶⁾ although a variety of factors pared the numbers down considerably. The study consisted of two distinct parts, one testing the drug on "mildly" depressed patients and the other "moderately" depressed patients. The following analysis is taken from the Food and Drug Administration "Review and Evaluation of Clinical Data: Amendment," dated December 30, 1985.

On various measurements, including the patients' ratings of themselves, the slightly larger group diagnosed as mildly depressed showed no improvement on Prozac - an observation that becomes more interesting in light of currently publicized claims that Prozac is especially helpful to mildly depressed people. The moderately depressed group included only 171 Prozac patients who actually completed the six-week trial. The moderately depressed patients did show some improvement, including on their subjective rating of themselves. There were no seriously depressed patients in the protocol.

In Protocol 62, the dropout rates were high in all dose ranges: over 35 percent at 20 mgs., over 40 percent at 40 mgs., and over 50 percent at 60 mgs. The most common reason for dropping out was adverse side effects and the second was lack of effectiveness. The most frequent side effects were nervousness, anxiety, insomnia, nausea, anorexia, and diarrhea, each of which occurred in more than 15 percent of patients. Even at the lowest dose, more than 70 percent of the patients experienced at least one side effect, and at the highest dose, 90 percent endured one or more.

The FDA is very critical of this protocol. It was unable to identify a specific "time period when the results became significant." There were no significant differences between the drug and the placebo before the fourth week. According to the December 30, 1985, FDA evaluation, this study was so "seriously flawed" in design that interpretation was difficult (pp. 12 and 13). The concluding line of the FDA's amended analysis states, "It is not- possible to arrive at a single, unequivocal interpretation of the results." Keep in mind that this protocol, like the others, used the placebo-washout strategy that skews the results in favor of the drug, and that this protocol was one of the three most positive in a much larger field of negative protocols.

As already noted in regard to Protocol 27, both Feighner and Cohn were seriously criticized by the FDA on various grounds. Both were principal investigators in Protocol 62. Fabre, whose legal situation we discussed in regard to Protocol 19, was also a principal investigator in Protocol 62.

Fieve Presents a Different View of Protocol 62

One of Lilly's hand-picked principal investigators in the ten-site Protocol 62 was psychiatrist Ronald R. Fieve, a professor at Columbia University and chief of psychiatric research at the New York State Psychiatric Institute. Fieve is the author of the 1994 mass-market book Prozac: Questions and Answers for Patients, Family, and Physicians.

In the preface to his book, Fieve describes how eager he was to participate as one of Eli Lilly's principal investigators and he tells the reader that the protocol did obtain the anticipated positive result for Prozac. But his analysis of Protocol 62 bears little resemblance to the facts on several critical points:

Fieve says that the protocol included a comparison antidepressant, Tofranil, but it did not."

Fieve states that the "pooled data" showed that Prozac worked better than placebo. As noted, there were two studies, and the pooled data for the mildly depressed patients showed no efficacy. A positive effect was barely measured only for the moderately depressed patients, a factor Fieve does not share with the reader. And, of course, there were no seriously depressed patients in the study, another fact Fieve does not share with the reader.

Fieve states that "Prozac patients experienced a startling absence of side effects compared to imipramine patients." Not only were there no imipramine patients for comparison, adverse side effects afflicted the vast majority of Prozac patients and helped to cause a very substantial drop-out rate.

Fieve says that Prozac was proven as effective as the older antidepressant, Tofranil, when the latter wasn't even included.

Elsewhere in his book, without citing the FDA studies, Fieve states "approximately 65% to 70% of depressed patients who take their Prozac are fully relieved within two to six weeks...." But he couldn't have concluded this from Protocol 62 or from any other Lilly-sponsored FDA study of Prozac. The high drop-out rates in all those research projects make it impossible to conclude that 65 to 70 percent of the patients were improved. At least one-third of the patients usually failed to finish the trials and drop-out rates sometimes reach 50 percent. Besides, the vast majority of protocols and individual studies showed no positive effect from Prozac.

In a telephone interview on April 6, 1994, I asked Fieve to clarify the discrepancies. Initially, Fieve thought he recalled an antidepressant control group in his Lilly-sponsored FDA study; but after taking a moment to check with his assistant, he acknowledged that his memory had been inaccurate. He thanked me and said that he would fix the error in the next edition of his book.

In the interview, Fieve repeated that other FDA studies did prove Prozac to be as effective as the older antidepressants, and was surprised when I assured him that in fact Prozac consistently proved less effective according to the FDA. Fieve's misperception about this is understandable, since Lilly's (as opposed to the FDA's) published version of Protocol 27, as we have already noted, claimed that the drugs did show comparable efficacy. Almost no one reads the FDA's analysis, which must be obtained through the Freedom of Information Act.

When I reminded Fieve, he acknowledged that his Lilly-sponsored Prozac study showed no positive effect on more than half the patients in his study - those diagnosed as mildly depressed - and that the FDA found the whole study very flawed.

I asked Fieve how he could write in his book that up to 70 percent of patients will improve on Prozac when the FDA studies showed that up to 50 percent were likely to drop out due to adverse effects and lack of efficacy. The numbers were contradictory.

Fieve explained that the 70 percent improvement rate cited in his book was based on his personal impressions from his extensive clinical practice in which he gives Prozac and other SSRIs (selective serotonin reuptake inhibitors) to most of his depressed patients. When pressed, he said that the 70 percent estimate did not include the one-third of his Prozac patients who dropped out, mainly due to adverse side effects. He explained that the 70 percent improvement rate was based on those patients who remain on the drug, not on all the patients who start taking the drug. But simple math shows that's an improvement rate of only 47 percent - less than placebo in many studies.⁽⁷⁾

Prozac Plus?

In chapter 4, we shall find that stimulant effects - including insomnia, nightmares, agitation, anxiety, and nervousness - are very commonly caused by SSRIs, especially Prozac. To counteract these adverse effects, clinicians frequently prescribe an additional drug along with Prozac. These include sleeping pills, such as chloral hydrate, or benzodiazepines (minor tranquilizers), such as Klonopin, Dalmane, or Xanax. Sometimes the sedative is prescribed with the first dose of the SSRI, sometimes it is prescribed only after stimulant symptoms, such as insomnia or agitation, begin to appear. Either way, the patient is exposed not only to the hazards of SSRIs but to those of sleeping pills and minor tranquilizers, which include addiction, withdrawal, and mental dysfunction.

In setting up and carrying out its protocols, Lilly seemed to recognize that some Prozac patients would also need to be put on sedatives or tranquilizers. According to the FDA's March 28, 1985, report, Protocol 27 states that patients will be excluded from the study if they take psychotropic drugs "other than benzodiazepines or chloral hydrate" (p. 5). Under "concomitant medications," the protocol seems to set a more narrow spectrum of permissible sedatives, stating that the only allowable medications were choral hydrate or flurazepam (Dalmane) for sleep.

In Protocol 19, patients were supposed to be excluded if they took any other psychoactive drugs (p. 52). However, there is a seeming contradiction, because under concomitant medications, it is stated that "According to the protocol, the only allowable concomitant medication was chloral hydrate for sleep and benzodiazepines (not further specified) for agitation" (p. 54).

In Protocol 62, as described in the FDA's December 30, 1985, report, patients were supposed to be excluded if they took any additional drug except chloral hydrate. However, in "actual practice" patients were not excluded if they took benzodiazepines (p. 2).

It is obviously important to know how many of the patients taking Prozac in the FDA trials were in fact taking Prozac plus a sedative, yet the numbers involved are not reported in Protocols 19, 27, or 62. However, during one of the attempts to pool the data for Protocol 27, Lilly decided to exclude all patients who took other drugs. According to the FDA's March 28, 1985, report, it turned out that patients were taking a wide variety of sedative drugs, including sedative tricyclic antidepressants, phenobarbital, and benzodiazepines (p. 48). Not counting the Cohn study, which had been dropped, 25 percent (135 of 540) of the enrolled patients were taking an additional drug.

Of critical importance is that when patients taking sedative drugs were removed from the pooled data, Prozac failed to show significant efficacy. Prozac proved efficacious only when patients taking the additional sedative drugs were reincluded in the pool (p. 50-A).

The inclusion of sedative drugs in Lilly's approval studies completely distorts them. In effect, the FDA ended up approving Prozac in combination with sedatives.

Prozac's Tests Show it to Be Anything But a Miracle Cure

To sum up, the three protocols we have examined were the only ones that the FDA judged valid enough to use for demonstrating efficacy. A large number of other studies were even more scientifically questionable or showed Prozac to be ineffective. These three badly flawed efforts, with many patients suffering adverse reactions, are the basis for the FDA allowing Prozac to be given to millions of Americans. A lot of fancy numbers-crunching was required to make Prozac look any better than a lowly sugar pill. In addition, several of the investigators were severely criticized by the FDA for their practices, including failure to observe protocol rules.

It bears repeating:

These three protocols - with only 286 Prozac patients finishing the four- to six-week studies - were the best that Eli Lilly and the FDA could come up with to prove the value of Prozac.

Is There Any Such Drug As An Antidepressant?

Recently the whole idea of an antidepressant drug has been challenged. If you believe, as we do, that depression is a form of psychological suffering based on hopelessness and despair, it would seem unlikely that drugging the brain could substantially help. On the other hand, any drug, including a sugar pill, might help by providing hope and encouragement, especially if the doctor offers it with authority and enthusiasm, and even more especially if the media are pushing it as well.

These observations have been recently confirmed in a series of new studies by psychologists Roger Greenberg, Ph.D., and Seymour Fisher, Ph.D. Both Greenberg and Fisher are professors at the Department of Psychiatry and Behavioral Sciences of the State University of New York Health Sciences Center in Syracuse." In 1992 a team led by Greenberg performed a statistical meta-analysis of 22 independent double-blind studies of three relatively new antidepressant drugs-Asendin (amoxapine), Ludiomil (maprotiline), and Desyrel (trazodone). Each study included both a placebo group and a comparison to an older drug.

According to self-ratings by the patients in these studies, the newer and older antidepressants were no better than placebo. In the words of Greenberg and his research team, "Thus, patient ratings of outcome for both standard and new antidepressant medication showed virtually no benefit beyond that obtained from placebo."⁽⁸⁾

While clinician ratings of patient improvement were a little more positive, Greenberg's research casts serious doubt on these evaluations. Studies show that clinicians are often able to guess who is receiving the drug and who is getting the sugar pill. Patients can also be influenced by side effects that inform them whether or not they are ingesting an active substance. They naturally feel more improved by an agent that has solid side effects.

To make a fair comparison between a drug and a placebo, the sugar pill needs a little punch of its own, something to make the patient think a real treatment is taking place. When patients are given active or enhanced placebos - drugs, such as atropine, that have side effects but no antidepressant effect - the patients improve as much as if they were on so-called antidepressants. In a 1993 report, Fisher and Greenberg concluded that it's time to view with skepticism any study in which the placebo does not have significant side effects.⁽⁹⁾

In discussing the study, Greenberg told Bruce Bower of Science News that an unpublished meta-analysis by himself and Fisher showed Prozac to be no better than any of the other antidepressant drugs, all of which had doubtful efficacy.

Despite the current enthusiasm for Prozac, the FDA studies underscored the drug's lack of effectiveness, and recent analyses of literature indicate that antidepressants in general are no better than placebo. But placebo, as Fisher and Greenberg reconfirmed, is very powerful. The lowly sugar pill - utterly free of brain-disabling and life-threatening adverse effects - improves a large portion of depressed patients.⁽¹⁰⁾ It may be the true miracle.

If Prozac isn't very effective, is it at least relatively safe? That was the hope within psychiatry when Prozac gained FDA approval. If it wasn't anything special therapeutically, it might have fewer or less serious side effects than other antidepressants. In the next three chapters, we turn to adverse reactions caused by Prozac, a subject of ultimately greater concern to consumers than Prozac's ineffectiveness as an antidepressant.

Footnotes:

1. In the final chapter we will look in greater depth at the question "What is depression?"

2. For the exclusion of "serious suicide risk" patients, see Food and Drug Administration, October 3, 1988, pp. 19, 21, and 23.

3. Now a full-time employee of the Upjohn Company, the maker of the psychiatric drugs Xanax and Halcion.

4. An October 3, 1988, Food and Drug Administration report states that 6,070 were exposed to Prozac in U.S. studies, and 1,850 abroad (p. 27).

5. Feighner was also involved in Protocol 62 (page 58) and it was unclear whether one or both Prozac protocols were involved in some of these criticisms.

6. The principal investigators were Roland J. Branconnier, Ph.D., and Eric Dessain, M.D.; Jay B. Cohn, M.D., Ph.D.-, M. Lynn Crimson, Pharm.D., and Allen Childs, M.D.; David L. Dunner, M.D.; Louis F. Fabre, Jr., M.D., Ph.D.; John P. Feighner, M.D.; Roland R. Fieve, M.D.; Joseph Mendels, M.D.; Ram K. Shrivastava, M.D.; Ward T. Smith, M.D.

7. But even these relatively unimpressive improvement rates cannot be taken too seriously, because clinical impressions from one's own practice are easily distorted to meet personal expectations or hopes. That's why controlled studies are required. It is not sound to publish subjective personal estimates in a book as if they have scientific validity. It is especially inappropriate not to delineate between clinical impression and empirical data when it comes to providing scientific-sounding percentages.

8. In contrast, Greenberg finds that studies of psychotherapy for depression demonstrate improvement on both patient and clinician ratings.

9. I came to the same conclusion in Toxic Psychiatry (1991).

10. Patients do frequently report side effects when taking placebo, but however real the effects seem, they do not cause actual brain dysfunction and damage or threaten the individual's life.

Peter Breggin's Home Site - Peter R. Breggin, M.D. founded The International Center for the Study of Psychiatry and Psychology (ICSPP) as a nonprofit research and educational network concerned with the impact of mental health theory and practices upon individual well-being, personal freedom, and family and community values. For 25 years ICSPP has been informing the professions, media and the public about the potential dangers of drugs, electroshock, psychosurgery, and the biological theories of psychiatry.

Suggested Reading:

Brain-Disabling Treatments in Psychiatry : Drugs, Electroshock, and the Role of the FDA Today! by Peter R. Breggin, M.D.

Toxic Psychiatry : Why Therapy, Empathy, and Love Must Replace the Drugs, Electroshock, and Biochemical Theories of the New Psychiatry by Peter R. Breggin, M.D.

The Manufacture of Madness : A Comparative Study of the Inquisition and the Mental Health Movements by Thomas S. Szasz, M.D., Professor

Law, Liberty, and Psychiatry : An Inquiry into the Social Uses of Mental Health Practices by Thomas S. Szasz, M.D., Professor

Bedlam : Greed, Profiteering, and Fraud in a Mental Health System Gone Crazy by Joe Sharkey

The Limits of Biological Treatments for Psychological Distress by Seymour Fisher and Roger P. Greenberg

Physician's Desk Reference (PDR)

Psychiatric Drugs: Hazards to the Brain by Peter R. Breggin, M.D.

Say NO To Psychiatry!

Back to Psychiatric Drugs Main Page

Back to Main SNTP Page

Pursuing Truth in all subjects...

©Gene Zimmer 1999 ALL RIGHTS RESERVED